| |

Received: 30 July 2018    Revised: 2 February 2019    Accepted: 2 February 2019

DOI: 10.1111/clr.13415

ORIGINAL RESEARCH

Comparison of three different types of implant-supported fixed

dental prostheses: A long‐term retrospective study of clinical
outcomes and cost‐effectiveness

Andrea Ravidà1  | Mustafa Tattan2  | Houssam Askar1  | Shayan Barootchi1  |

1 1
Lorenzo Tavelli  | Hom‐Lay Wang

1
Department of Periodontics and Oral
Medicine, University of Michigan School of
Dentistry, Ann Arbor, Michigan
2
Department of Periodontics and
Iowa Institute for Oral Health
Research, University of Iowa College of
Dentistry, Iowa City, Iowa
Correspondence
Hom‐Lay Wang, Department of Periodontics
and Oral Medicine, University of Michigan
School of Dentistry, Ann Arbor, MI.
Email: homlay@umich.edu
Abstract
Objective: To study the performance of 2–3 posterior bone‐level dental implants
constructed with either three non‐splinted crowns (NSC), three splinted crowns (SC),
or a 3‐unit implant‐supported bridge over two implants (ISB).
Material and methods: Patients treated with three metal‐ceramic NSC, SC, or an ISB
were included in the present retrospective study. Implant survival and success rate as
well as all biological and technical complications were collected. The cost associated
with each of the treatment options was evaluated in the comparative analysis.
Results: One hundred and forty‐five patients (40 NSC, 52 SC, and 53 in the ISB) re‐
ceiving 382 bone‐level implants (120 NSC, 106 ISB, and 156 SC) were included (mean
follow‐up of 76.2 months). Lack of success was observed in 33.8% of the total patient
sample, being lower in the ISB group. Implant survival rates were 92.5% in the NSC,
100% in the ISB, and 88.5% in the SC, with significant difference noted between the
ISB and SC (p = 0.01). Overall, 9.9% of the total implants were found to have peri‐im‐
plantitis (PI), with 16.7% in the SC, 7.5% in the NSC, and 2.8% in the ISB. Patients
presenting prosthodontic complications were significantly higher in NSC (32.5%)
than ISB (13.2%) and SC (15.4%). The total cost of the ISB group was significantly
lower when compared to the NSC and SC groups (p < 0.001).
Conclusions: An 3‐unit implant‐supported bridge restoring 2 implants seems to pre‐
sent the most ideal long‐term therapeutic solution, among the investigated ap‐
proaches in this study, in rehabilitating a 3‐unit edentulous area.

KEYWORDS
dental implants, implants, implant‐supported fixed dental prosthesis, partially edentulous,
splinted

1 |  I NTRO D U C TI O N
Dental implants have played an integral role in the management of
partially edentulous patients in both the maxilla and mandible (Jung,
Zembic, Pjetursson, Zwahlen, & Thoma, 2012). Although long‐term
success has been widely established, the outcome is influenced by
factors that guide clinicians to select the most appropriate surgical
and prosthodontic approach (Flemmig & Beikler, 2009; Grossmann,
Finger, & Block, 2005).
Among these factors is determining whether to use a prostho‐
dontic component composed of splinted versus non‐splinted crowns
(Ravida, Barootchi et al., 2018). Splinted crowns tend to distribute

Clin Oral Impl Res. 2019;30:295–305. wileyonlinelibrary.com/journal/clr   © 2019 John Wiley & Sons A/S. |  295
Published by John Wiley & Sons Ltd
 |
296       RAVIDÀ et al.
the occlusal forces placed on the implants, resulting in a less fre‐
quent occurrence of prosthodontic complications and decreased
strain on the surrounding peri‐implant bone (Clelland, Chaudhry,
Rashid, & McGlumphy, 2016; Yilmaz, Seidt, McGlumphy, & Clelland,
2011). However, the literature supporting this treatment option is
often conducted through finite and photoelastic analyses due to
the ethical boundaries preventing the use of occlusal overload in
human subjects (Guichet, Yoshinobu, & Caputo, 2002; Huang et al.,
2005; Yoda et al., 2016). In 2002, Guichet et al. demonstrated a re‐
duced overall peak stress induction around the central implant of
a 3‐unit splinted crown restoration, whereas the stresses were con‐
centrated around all the loaded implants when not splinted (Guichet
et al., 2002). Similar results were also reported by Nissan, Ghelfan,
Gross, and Chaushu (2010) where less load to the crown margin was
observed with splinted versus non‐splinted implant restorations
(Nissan et al., 2010). Although these arguments present advantages
for splinted restorations, maintaining adequate oral hygiene within
the interproximal spaces is an essential practice to avoid the inci‐
dence of peri‐implantitis (PI) (Serino & Strom, 2009). This would
render three non‐splinted crowns an advantageous option over
splinted crowns as a prosthodontic approach, particularly in patients
with history of periodontitis and/or limited dexterity when cleaning
(Renvert & Persson, 2009). An additional disadvantage of splinting
implant‐supported crowns is the challenge of framework fit and pro‐
viding an adequate emergence profile (Ravida, Saleh, Muriel, Maska,
& Wang, 2018).
Another determining factor is the number of implants required
to rehabilitate a partially edentulous area. This presents a conflict
between one implant per missing tooth and an implant‐retained
bridge (Eliasson, Eriksson, Johansson, & Wennerberg, 2006). The
use of a single implant per lost tooth seems to pose a plausible
clinical choice in the reduction of specific risk factors such as over‐
load (de Souza Batista et al., 2017). However, several investigations
have demonstrated successful full‐arch rehabilitation via cross‐arch
splinted prostheses supported by a fewer number of implants than
lost teeth (Cannizzaro et al., 2011; Malo, Nobre, & Lopes, 2013;
Malo, Rangert, & Nobre, 2003). In the treatment planning of restor‐
ing a 3‐unit edentulous area, the lack of space and poor bone quality
may interfere with the use of three implants. Such a limitation can
be overcome by using two implants supported by a bridge (de Souza
Batista et al., 2017). Furthermore, an often‐overlooked influencing
factor is the cost (Ravida, Barootchi et al., 2018). The use of two
versus three implants could likely influence the clinical decision of
therapy. However, it is essential to compare the total cumulative cost
including all the potential complications that may accompany each
therapeutic option.
Literature comparing the rehabilitation of 3‐unit edentulous
areas in the posterior maxilla or mandible using two versus three
implants remains scarce (Eliasson et al., 2006; Yi, Lee, & Kim, 2013).
In addition, no previous article has evaluated the cost‐effective‐
ness of each therapeutic approach yet. Therefore, the aim of this
study was to compare the survival rates, success and prosthodontic
complication rates, incidence of PI, and cost‐effectiveness of the
three clinical options considered in rehabilitating a 3‐unit edentulous
area in the posterior maxilla or mandible.
2 | M ATE R I A L S A N D M E TH O DS
This study was approved by the University of Michigan, School
of Dentistry, Institutional Review Board for Human Studies
(HUM00114380). This retrospective investigation included all pa‐
tients treated with two to three bone‐level implants (loaded with
either three non‐splinted crowns, three splinted crowns, or an im‐
plant‐supported bridge) (Figure 1) restoring a posterior 3‐unit eden‐
tulous area between January 1990 and September 2017 at the
University of Michigan School of Dentistry, Ann Arbor, MI, USA.
The physical and digital records that fall under the predetermined
eligibility criteria were screened and evaluated by three examiners
(MT, AR, SA). Any disagreement that arose during the evaluation and
data collection process was resolved through discussion with the su‐
pervising investigator (HLW).
2.1 | Inclusion criteria
1. Partially edentulous patients treated with 2–3 implants restoring
a posterior (molars and premolars) maxillary or mandibular 3‐unit
edentulous area with a documented follow‐up of ≥1‐year fol‐
lowing implant loading.
2. Cases with all dental implants placed during the same surgical
appointment.
3. Patients who received implants loaded with titanium prefabri‐
cated abutments and either three metal‐ceramic splinted crowns
(SC), three non‐splinted metal‐ceramic crowns (NSC) or an im‐
plant‐supported bridge (ISB) over two implants.
4. The presence of opposing occlusion (teeth/Implants).
2.2 | Exclusion criteria
1. Partially edentulous patients who have received any of the
aforementioned choices of prosthodontic rehabilitation methods
loaded on >3 implants.
2. Patients with restored 3‐unit edentulous areas in the anterior
zone.
3. Patients with ambiguous or incomplete charts.
4. Patients with a <1‐year follow‐up period.
5. Medically compromised patients (any past records of uncon‐
trolled diabetes, radiation and/or chemotherapy treatment, psy‐
chological problems) and severe bruxism cases (diagnosed and/or
self‐reported).
6. Patients treated or maintained in centers outside the University
of Michigan School of Dentistry.
7. Patients with inaccessible files due to bad debt, destroyed record,
or decease.
RAVIDÀ et al.
2.3 | Data collection and classification
All patient records were initially screened and evaluated against the
aforementioned eligibility criteria. Subsequently, the selected cases
were separated into three‐study groups based upon the selected
definitive prosthesis: non‐splinted crowns group, splinted crowns
group, and fixed partial denture group (Figure 1).
As part of the data collection process, all relevant patient in‐
formation, including age (at the time of implant placement), gender,
presence of a smoking habit (≥1 cigarette/day), diabetes (validated
via the patient's medical records) and history of periodontal dis‐
ease, was obtained. A positive history of periodontal disease was
determined to be present if ≥4 sites presented with ≥3 mm clinical
F I G U R E 1   Clinical depiction of the three different treatment
approaches represented in the study groups; (1) three posteriors
implants loaded with three non‐splinted crowns (NSC); (2) two
posterior implants loaded with an implant-supported bridge (ISB);
(3) three posteriors implants loaded with three splinted crowns (SC)
|
      297
attachment loss, and there was a past history of scaling and root
planing (Armitage, 2004; Pihlstrom, Michalowicz, & Johnson, 2005),
based on each patient's documented periodontal charts. Additional
data including the number of implants and their positions, implant
description (brand, length, and diameter), and the type of crown re‐
tention (cement or screw retention) were also collected.
2.4 | Peri‐implantitis, survival, and success rate
The definition for PI proposed by the 8th European Workshop on
Periodontology in 2011 (Tonetti, Palmer, & Working Group 2 of the
VIII European Workshop on Periodontology, 2012), where PI was
defined as clinical inflammation in combination with radiographic
marginal bone loss >2 mm, was used to classify cases into positive or
negative for PI. Using a commercially available software (ImageJ, U.
S. National Institutes of Health, Bethesda, Maryland, USA), marginal
bone levels as well as the horizontal and vertical distance between
the implants of an individual case were measured at baseline (time
of prosthodontic loading). Meanwhile, the degree of marginal bone
loss associated with each included implant was recorded at the final
follow‐up appointment utilizing calibrated periapical radiographs
via the same software (Schneider, Rasband, & Eliceiri, 2012). Two
individual calibrated examiners (AR & SB) performed the calcula‐
tions separately, but if significant differences arose, a third reviewer
(MT) was included for reassessing the radiographs to arrive at a final
resolution. Peri‐implantitis was first evaluated per patient, then
per implant individually. The prevalence of PI was recorded using
a binary score for each implant (0 for a healthy implant, 1 for a dis‐
eased implant) to calculate the percentage of diseased implants. To
collect data on each implant position independently, each implant
was assigned an identification alphabet based on location. This in‐
cluded labeling all mesial implants as “Implant A,” all central implants
as “Implant B” and all distal implants as “Implant C.” Similar values
were assigned to patients based on the presentation of PI around
any implants (0 for a patient with all healthy implants, 1 for a patient
with radiographic signs of ≥1 diseased implant). Implant failure was
defined as a removed, lost, mobile, or fractured implant (Chrcanovic,
Albrektsson, & Wennerberg, 2014), and calculated for each implant
individually and then for each patient with the same standards used
previously for PI. Success rate was evaluated at the patient level,
where a case (SC, NSC or ISB) was classified as successful when an
absence of biological and prosthodontic complication throughout
the follow‐up period was demonstrated (0 for a patient with some
prosthodontic or biologic complication, 1 for a patient without any
prosthodontic or biologic complication).
2.5 | Prosthodontic complications
Prosthodontic complications included a (a) fractured/chipped/re‐
placed prosthesis, (b) fractured prosthesis, (c) crown/prosthesis de‐
cementation, and/or (d) loosened abutment screw, and along with
the associated management were registered at the patient follow‐up
appointments.
 |
298       RAVIDÀ et al.

provided treatments, actual cost was presumed within the particular

2.6 | Case follow‐up periods
patient's cost of treatment.
To ensure more meticulous data analysis, three independently de‐ Within the initial cost, every treatment fee, such as preliminary
fined follow‐up periods were recorded during data acquisition. These consultation appointments, use of radiographic and/or laboratory
were (a) follow‐up based on implant survival, (b) follow‐up based on diagnostic aids, laboratory fees and preparations, and the entire cost
implant loading (prosthodontic follow‐up), and (c) follow‐up based of surgery, was included. Complication management cost included
on the occurrence of PI. The marked period for the follow‐up based any fee related to follow‐up maintenance, as well as management of
on implant survival was the duration between implant placement any complication pertaining to any component of the implant pros‐
and final documented date during which the implant remained in the thesis structure.
oral cavity. In the case of an implant having been lost or extracted, The average cost of each procedure was calculated as follows:
the date of disease presentation was decided as the final follow‐up
cost1 + cost2 + cost3 + cost4 … ∕n
mark. The set period for the follow‐up based on implant loading was
the duration between implant loading and final documented clinical
where,
appointment date. At any point during the prosthodontic follow‐up
Costx = Procedure Cost at a Given Year.
period where a prosthodontic complication arose, the timeframe (in
n = Total number of Costx events per procedure.
months) was recorded for data analysis. Finally, the marked period
for the follow‐up based on the PI was the duration between implant
placement and the last radiograph in which the bone around the im‐ 2.8 | Statistical analyses
plants was visible.
The demographic profile, clinical characteristics, and post hoc
power analysis of the included sample were compared using: (a)
2.7 | Cost t test analyses (t); (b) Chi2 homogeneity tests (Chi2); (c) ANOVA
F test (F); (d) Mann–Whitney test (MW); and (e) Kruskal–Wallis
The cost analysis was patient centered aimed at distinguishing all the
test (KW). The probability of PI, implant failure, and prosthodon‐
costs of diagnostic, therapeutic, and follow‐up procedures. The pri‐
tic complications among the three groups was calculated using a
mary objective of this analysis was to achieve a holistic comprehen‐
multiple binary logistic regression: estimation of odds ratio (OR)
sion of the cost‐effectiveness related to each of the three treatment
adjusted by smoking, history of periodontitis, follow‐up time, gen‐
modalities studied in this investigation, as well as their entailing
der, age, diabetes, arch, and bone augmentation. The McNemar
complications.
test was performed to compare the incidence of PI in two spe‐
The average cost of individual clinical procedures across the 1‐ to
cific positions of the same patient. Peri‐implantitis at the implant
22‐year follow‐up period (upper limit determined by the case with
level was estimated using a generalized estimation equation (GEE)
the longest follow‐up) was predetermined, to ensure standardization
model to determine intra‐subject correlations. The effect of the
among the sample, and utilized in the analysis. This was performed
horizontal/vertical distance on the PI probability was evaluated
by calculating the mean of the individual costs of a procedure
by incorporating this covariate to the previous model. The survival
from every year since 1997, at the University of Michigan School
rates of the three groups were analyzed and compared using by
of Dentistry. This precludes any interference, by the regular rate of
Chi2 and Fisher's exact test. At the implant level, to evaluate the
inflation across the 1‐ to 22‐year timeline, with the reliability of the
survival rate, the Kaplan–Meier function was performed, and the
cost analysis, standardizing the cost for all patients. A pricelist was
Log Rank (Mantel‐Cox) test with hazard ratio estimation was used
generated based on these means, and all procedures associated with
for the overall comparison of the survival curves. Finally, the cost
each patient file were recorded by three study investigators (MT, AR,
analyses were performed using non‐parametric tests of Kruskal–
SA). Wherever doubt arose, the supervising investigator (HLW) was
Wallis and Mann–Whitney. The level of significance used in the
referred to. With these records, the cost of treatment and manage‐
analyses was set at 5% (α = 0.05).
ment performed on each patient was noted and computed into the
following categories of cost:

1. Initial Cost: Implant +Prosthesis Placement Fees
2. Cost of Complication Management: Prosthodontic Complication
Management Fees
3. Total Cost: Initial Cost +Cost of Complication Management
The purpose of analysis was to simulate a clinical setting where
a patient is not pardoned for payment as a means to have a fair
and elaborate comparison between the treatment approaches.
Therefore, whether or not the patient had actually paid for the
3 | R E S U LT S
3.1 | Clinical characteristics and demographic
profiles
A total of 329 3‐unit posterior edentulous sites restored with an
implant‐supported fixed dental prosthesis were initially screened
for possible inclusion. Following thorough examination, 145
cases (pertaining to 64 males and 81 females with a mean age of
60.7 ± 10.1 years old) were selected and subsequently divided into
RAVIDÀ et al.
F I G U R E 2   Percentage distribution of treatment modalities
within the study sample
three study groups (40 in the NSC, 52 in the SC, and 53 in the ISB)
(Figure 2) for analysis. The remaining cases were excluded based on
the following reasons: (a) <1‐year follow‐up (52), (b) anterior zone
cases (32), (c) ineligible number of implants or final prosthodontic
designs (32), (d) files with incomplete clinical information (16), (e) files
with no relevant cases (15), (f) implant placement or loading (single
crown group) occurred in different appointments (14), (g) implant
loading performed in centers outside the University of Michigan
School of Dentistry clinics (14), (h) inaccessible files (8), or (i) patients
with uncontrolled diabetes (1).
Overall, 382 implants (120 in the NSC, 106 in the ISB, and 156
in the SC) were included in the present study. Table 1 provides the
demographic and baseline clinical parameters. All prostheses were
porcelain‐fused‐to‐metal and either screw retained (standard pro‐
tocol for prosthetic screw tightening via a torque controller set
at 30 Ncm) or cement retained using premier implant cement kit
(Premier Dental, Plymouth Meeting, PA, USA). Patient age, implant
location (maxillary or mandibular), and bone augmentation were
found to be significantly different among the three groups (p < 0.05).
A general linear model type ANOVA, with the current sample
of n = 145 patients, reached a power of 76.4% to detect an average
effect size (f = 0.25) in the mean MBL difference between the three
types of prostheses as significant, assuming a 95% confidence level.
The same model at the implant level (n = 382) achieved a 90.9%
power with the same conditions. This value was corrected by the
dependence between observations (multiplicity of implants per pa‐
tient), assuming a moderate correlation (ρ = 0.5).
3.2 | Survival rate
At the patient level, 93.8% of the patients did not experience im‐
plant failure. When this percentage was equated among the studied
groups, survival rates were 92.5% in the NSC, 100% in the ISB and
88.5% in the SC (Figure 3a). An inter‐group comparison showed a sta‐
tistically significant difference between the ISB and SC (p = 0.01), in
addition to a trend toward less failure in ISB when compared to NSC
(p = 0.07). The difference between the NSC and SC was not statisti‐
cally significant (p = 0.58). At the implant level, 366 (95.8%) of the
total 382 implants were still in function by the end of the total study
|
      299
follow‐up period (97.5% in the NSC, 100% in the ISB, and 91.7% in
the SC). A statistically significant difference between the ISB and SC
(p < 0.01) was also demonstrated at this level. However, as opposed
to the patient level, a statistical trend was observed between the
NSC and SC (p = 0.04) at the implant level, while ISB reported again
less failure rate than SC (p = 0.07). The Log Rank (Mantel‐Cox) test
confirmed the non‐equality in survival distribution between the ISB
and SC (p < 0.01) (Figure 3b).
3.3 | Success rate
The total success rate in the entire patient sample was 66.2%. This
percentage was distributed across the study groups as 81.1% in the
ISB, 61.5% in the SC, and 52.5% in the NSC (corresponding success
rate percentages summarized in Figure 4). Patients with an ISB com‐
pared to NSC demonstrated a decreased probability of implant‐re‐
lated complications of up to 74% (OR = 0.26, p = 0.004). This can
also be viewed as NSC being at a 289% greater risk of implant‐re‐
lated complications than an ISB (OR = 3.89, p = 0.004). However, the
risk between SC and NSC was comparable (p = 0.385). Similarly, SC
is at a 168% higher risk of developing implant‐related complications
than an ISB (OR = 2.69, p = 0.029).
3.4 | Peri‐implantitis
A total of 16.6% of the patient sample developed PI (in ≥1 of their
implants). This percentage fluctuates from 5.7% in the ISB to 17.5%
in the NSC and 26.9% in the SC. Table 2 outlines the same infor‐
mation as the number of affected implants in the same individuals.
Overall, 9.9% of the total implants were found to have PI, with 16.7%
in the SC, 7.5% in the NSC, and 2.8% in the ISB (Figure 5a). Hence, a
patient with an ISB has a decreased probability of developing PI by
up to 72% (OR = 0.28, p = 0.082) when compared to NSC. Similarly,
a patient with SC was shown to be at an increased risk of develop‐
ing PI by about 74%, when compared to NSC (OR = 1.74, p = 0.289).
Finally, a patient with an ISB presents a significantly reduced risk
(84%) of developing PI as compared to SC (OR = 0.16, p = 0.007). At
the implant level, an implant under an ISB has a decreased probabil‐
ity of developing PI by up to 64% (OR = 0.36, p = 0.090) with respect
to being under SC. In the same way, an implant splinted to a second
implant demonstrated 2.5 times (or 147%) greater risk of acquiring
PI than that associated with single crown prostheses (OR = 2.47,
p = 0.156). Finally, an implant associated with an ISB showed to be at
a significantly reduced risk of developing PI (95.4%) as opposed that
having a splinted prosthesis (OR = 0.146, p = 0.003).
Results from the GEE depicted that neither the horizontal dis‐
tance between the implants (p = 0.5) nor the vertical distance
(p = 0.4) were factors associated with the incidence of PI.
Figure 5b summarizes the PI rate the implants according to location
(A, B, and C) among the three groups. No differences in PI incidence
were detected between the mesial and distal implants (A and C, re‐
spectively) among the study groups, while the central implants (B) per‐
taining to the SC demonstrated a significantly higher probability of PI
 |
300       RAVIDÀ et al.

TA B L E 1   Profile of the treatment groups according to variables at the patient and implant level

Groups

Single ISB Splinted p‐Value

Patient level
N (patients) 40 53 52
Age (years) 58.0 ± 8.3 (56.5) 63.4 ± 10.8 (64.0) 60.2 ± 10.0 (60.5) 0.032*  (F)
Women 22 (55.0) 33 (62.3) 26 (50.0) 0.445 (χ2)
Smokers 10 (25.0) 13 (24.5) 16 (30.8) 0.733 (χ2)
Diabetes 3 (7.5) 3 (5.7) 4 (7.7) 0.905 (χ2)
History of periodontitis 14 (36.8) 9 (18.0) 19 (38.0) 0.057 (χ2)
Maxilla 22 (55.0) 26 (49.1) 16 (30.8) 0.045*  (χ2)
Mandible 18 (45.0) 27 (50.9) 36 (69.2)
Cement retained prosthesis 12 (30.0) 12 (22.7) 12 (23.1) 0.761 (χ2)
Screw retained prosthesis 28 (70.0) 41 (77.3) 40 (76.9)
Bone augmentation 8 (20.0) 13 (24.5) 26 (50.0) 0.003**  (χ2)
Guided implants 13 (32.5) 16 (30.2) 10 (19.2) 0.289 (χ2)
Implant failure 3 (7.5) 0 (0.0) 6 (11.5) 0.046*  (χ2)
Follow‐up (months) survival rate 85.2 ± 40.4 (78.5) 80.1 ± 51.8 (63.0) 92.1 ± 51.2 (85.0) 0.272 (KW)
Follow‐up (months) prosthodontic 72.6 ± 34.8 (70.0) 69.8 ± 51.7 (49.0) 85.4 ± 53.5 (79.5) 0.254 (KW)
complications
Follow‐up (months) PI 61.8 ± 37.4 (58.0) 64.9 ± 50.2 (43.0) 78.7 ± 46.2 (60.0) 0.605 (KW)
Prosthodontic complications 13 (32.5) 7 (13.2) 8 (15.4) 0.051 (χ2)
0.292 (χ2)* * * * 
De‐cementation 6 (15.0) 0 (0.0) 4 (7.7) 0.018* (χ2)
Chipping 6 (15.0) 5 (9.4) 4 (7.7) 0.505 (χ2)
Fracture 4 (10.0) 2 (3.8) 0 (0.0) 0.058 (χ2)
Screw loosening 1 (2.5) 0 (0.0) 0 (0.0) 0.269 (χ2)
Peri‐implantitis 7 (17.5) 3 (5.7) 14 (26.9) 0.025* (χ2)
0.056 (χ2)* * * * 
Success rate 21 (52.5) 43 (81.1) 32 (61.5) 0.013*  (χ2)
0.034* (χ2)* * * * 
Implant level
n (implants) 120 106 156
Implant failure 3 (2.5) 0 (0.0) 13 (8.3) <0.001***  (χ2)
PI 9 (7.5) 3 (2.8) 26 (16.7) 0.009* *  (χ2)
0.033*  (χ2)* * * * 
PI implants A 2 (5.0) 2 (3.8) 8 (15.4) 0.091 (χ2)
0.275 (χ2)* * * * 
PI implants B 2 (5.0) – 12 (23.1) 0.029*  (χ2)
0.042*  (χ2)* * * * 
PI implants C 5 (12.5) 1 (1.9) 6 (11.5) 0.180 (χ2)
0.266 (χ2)* * * * 
Horizontal distance A–B 3.14 ± 1.37 (2.84) 3.25 ± 1.04 (3.02) 0.733 (t)
Horizontal distance A–C 9.04 ± 2.37 (8.52) –
Horizontal distance B–C 3.34 ± 1.04 (3.14) 3.91 ± 1.20 (3.87) 0.062 (t)
Vertical distance A–B 0.95 ± 0.78 (0.76) 1.02 ± 0.83 (0.78) 0.738 (t)
Vertical distance A–C 1.13 ± 0.90 (0.93) –
Vertical distance B–C 0.97 ± 0.71 (0.88) 1.08 ± 1.06 (0.62) 0.676 (t)
2
Note. Numbers are expressed as: Number of cases (%) or mean ± standard deviation (median). Chi test results, F test of the ANOVA model, two‐sample
t test, non‐parametric Kruskal–Wallis test, Chi2 Wald model GEE test, logarithm Kaplan–Meier (LR) test.
PI: Peri‐implantitis.
*p < 0.05; **p < 0.01; ***p < 0.001; ****p‐value of the adjusted model.
RAVIDÀ et al. |
      301

F I G U R E 3   (a) Rates of implant failure at the implant and patient level across the study groups, (b) cumulative survival rates of each study
group sample

than those of the NSC (OR = 5.70, p = 0.029). The risk is quantified as a

470% greater risk between the two. Furthermore, when the compari‐
son among implants A, B, and C within the same prosthodontic design
was performed, only implant B within a splinted prosthodontic design
tends to present with a higher PI rate than implant C (p = 0.070).

3.5 | Prosthodontic complications
Table 3 gives an overview of all occurred complications with respect to F I G U R E 4   Success rates represented at the implant level of
time and the number of prostheses that were being followed. During the each study group samples
entire study period, prosthodontic complications were found in 19.3%
of the patients. NSC, ISB, and SC had 32.5%, 13.2%, and 15.4% pros‐ TA B L E 2   Number of implants with PI by group
thodontic complications, respectively (Figure 6a). A patient with an ISB
Group
demonstrated a reduced risk of complications by up to 68% (OR = 0.32,
p = 0.029) when compared to NSC. This can also be expressed as a Single
216% higher risk for NSC compared to an ISB (OR = 3.16, p = 0.029). Total crowns ISB Splinted

Furthermore, the SC group showed a reduced risk for prosthodontic N % N % N % N %

complications (OR = 0.38, p = 0.057) when compared to NSC, where
Total 145 100 40 100 53 100 52 100
NSC had a risk 165% higher than SC (OR = 2.65, p = 0.057). Additionally,
0 121 83.4 33 82.5 50 94.3 38 73.1
each added year reduced the risk of the overall complications by 6%
1 15 10.3 6 15 3 5.7 6 11.5
(OR = 0.94, p = 0.031). The different types of prosthodontic complica‐
2 4 2.8 0 0.0 0 0.0 4 7.7
tions per group are outlined in Figure 6b. Out of all the documented
complications, 10.3% had crown chipping, 6.9% de‐cementation, 4.1% 3 5 3.4 1 2.5 0 0.0 4 7.7

crown fracture, and 0.7% prosthodontic screw loosening.
When the different types of complications were compared, the ISB
were associated with less de‐cementation than the NSC (p = 0.005)
and SC (p = 0.057). With respect to the occurrence of crown fractures,
the SC portrayed a superior outcome to the NSC (p = 0.033).
ISB: implant-supported bridge.
ISB (p = 0.001) (Figure 7). Finally, the total cost of the ISB group was
significantly lower than the NSC and SC (p < 0.001). Although there
were no differences between the ISB and SC at the complication level
of analysis, the initial cost of the SC is higher than those of an ISB.

3.6 | Cost
As depicted in Table 4, the primary factor to be considered was the
initial cost associated with each prosthodontic design, where ISB
costs only $6,998 (−16% with respect to the other two prostheses).
The higher source of variability was in the costs per complication and
on average was higher for the NSC than the SC (p = 0.081) and the
4 | D I S CU S S I O N
The choice of therapeutic approach pertaining to the final restora‐
tion of a posterior 3‐unit edentulous site is at the discretion of the
operator, and it is a common occurrence that different clinicians tend
 |
302       RAVIDÀ et al.

by Eliasson and coworkers (Eliasson et al., 2006; Yi et al., 2013), since

F I G U R E 5   Peri‐implantitis incidence rates, (a) at the patient
and implant levels and (b) according to implant position, across the
study sample
to lean toward certain approaches for the correction of the same
case. The present study presents a clear perspective of the perceiva‐
bly comparable treatment choice in the aforementioned clinical sce‐
nario, addressing all the clinical variables, including patient‐related
outcomes such as the associated finances, to provide clinicians with
objective criteria when selecting the most appropriate therapy. The
design of the present study is slightly different from that completed
we focused our attention on the posterior region and we included
three distinct study groups (SC, ISB, and NSC).
Our results suggest that the NSC and SC (three implant‐sup‐
ported prosthesis) had a greater incidence of PI than the ISB. This
implies that the higher bending moment associated with two im‐
plants in vitro (de Souza Batista et al., 2017) did not have a detrimen‐
tal impact on marginal bone loss. This was further confirmed by the
higher rate of PI in SC as opposed to an ISB (Hasan, Bourauel, Keilig,
Stark, & Luckerath, 2015; Huang et al., 2005; Shigemitsu et al., 2013;
Yilmaz et al., 2011).
Clinically, the present study does not reflect the results of a 10‐
year randomized controlled trial comparing three splinted versus
three non‐splinted implants in 44 patients (132 implants), where not
a single implant exhibited failure or biologic complications such as
PI or peri‐implant mucositis (Vigolo, Mutinelli, Zaccaria, & Stellini,
2015). This discrepancy between the two studies could be attributed
to: different treatment environments of academic setting versus
private practice; prospective versus retrospective study; and experi‐
enced surgeon versus beginner between their and ours, respectively.
A peculiar trend emerging from the analysis is that the central im‐
plant (B) under SC has a nearly sixfold higher probability of develop‐
ing PI than the central implant under NSC and higher PI rate than the
adjacent distal implant (C). Although the higher inter‐group difference
could be explained by the reduced cleanability of splinted prostheses,
the finding that within the same group the distal implant (often consid‐
ered more challenging to clean) is less affected presents an interesting
novelty that deserves more attention in the future studies.
In addition, residual cement may also be another contributing factor
for the incidence of PI (Wilson, 2009). However, in the present study, only
12 patients of each group's had received a cemented prosthesis (Table 1),
accounting to, at most, <30% of the patients in each set. Given that at

TA B L E 3   Incidence of prosthodontic complications per year for the three studied groups

Incidence of events (in months)

Type of Prosthodontic
prosthesis complication 0–12 12–24 24–36 36–48 48–60 60–120 120+ Total

ISB De‐cementation 0
Fracture/Chipping 4 2 1 1 8
Prosthodontic screw 0
loosening
No. of cases 53 53 48 33 27 26 7
SC De‐cementation 4 3 7
Fracture/Chipping 1 1 1 3 6
Prosthodontic screw 0
loosening
No. of cases 52 52 46 42 36 31 13
NSC De‐cementation 1 1 3 1 3 9
Fracture/Chipping 2 6 3 2 1 4 18
Prosthodontic screw 1 1
loosening
No. of cases 40 40 36 36 30 26 5

Note. ISB: implant‐supported bridge on two implants; NSC: three non‐splinted crowns; SC: three splinted crowns.
RAVIDÀ et al. |
      303

F I G U R E 6   Prosthodontic complication rates, (a) at the patient level and (b) according to type of complications, across the study sample

TA B L E 4   Cost expenditure through the

Groups
overall follow‐up time
Single crowns ISB Splinted p‐Value

Initial cost 8.301 6.998 8.301 –

Cost for 356.4 ± 925.8 95.5 ± 602.7 (0.0) 182.0 ± 692.4 0.005** 
complications (0.0) (0.0)
Total cost 8.657.4 ± 925.8 7.093.5 ± 602.7 8.483.0 ± 692.4 <0.001*** 
(8.301) (6.998) (8.301)
*p < 0.05, **p < 0.01; ***p < 0.001. ISB: implant-supported bridge.

F I G U R E 7   Projected comparison
of costs pertaining to each treatment
approach study group

baseline, the patient distribution in terms of prosthesis type heavily dis‐
favors a substantial presence of cement around the implants, and we felt
that it may be misleading stress the relevance of cement to the occur‐
rence of peri‐implant disease within the confines of this investigation.
The implant survival rate was calculated with and without the
Kaplan–Mayer test. When this test was not used, the rate was
calculated without considering that there were censored cases
throughout the observation period. The Kaplan–Mayer methodol‐
ogy allowed for the estimation of the survival function accounting
for the duration in which the implants were in the mouth. Hence,
the longer the follow‐up period, the lower the number of implants
considered. Thus, the 77% survival rate of the SC at the end of the
 |
304       RAVIDÀ et al.
follow‐up period is based on only a few cases that reached the final
follow‐up and should be interpreted with caution.
The high survival rate demonstrated by the ISB (100%) is not
striking. Yi and coworkers previously reported no failure in an av‐
erage follow‐up of 4 years (Yi et al., 2013), and similarly, Spies,
Witkowski, Butz, Vach, and Kohal (2016) confirmed these results
across a three‐year follow‐up period using all‐ceramic bi‐layered
implant‐supported 3‐unit fixed dental prostheses. Excellent results
(96.8% survival rate) have also been reported at 5 years of follow‐up
(Eliasson et al., 2006).
The ISB has also portrayed higher success rates than the other
groups, with 81% of patients without any reporting of complications
or concern throughout the entire follow‐up. Contrarily, despite the
higher PI rate and implant failures, SC presented a slightly higher suc‐
cess rate (61.5%) than NSC (52.5%). The explanation for this is the
higher resistance to prosthodontic complication that accompanies
the splinted crowns, particularly due to their resistance to rotational
movements enhancing their stability to eccentric forces (Faucher &
Bryant, 1983). In fact, NSC was the prosthodontic design that had pre‐
sented with more complications such as crown de‐cementation and
prosthodontic screw loosening. This observation has been previously
reported in a prospective split‐mouth study where splinted and non‐
splinted implant crowns (placed in 15 patients) were observed along
3 years, and all screw loosening reportedly only occurred on the non‐
splinted side (Clelland et al., 2016). Furthermore, a systematic review
found that screw loosening was associated with 5.6% of the implants
restored with splinted crowns and with 12.7% of those restored with
single crowns along a five‐year follow‐up period (Pjetursson, Bragger,
Lang, & Zwahlen, 2007). Also, similarly to our results, they reported a
higher incidence of ceramic chipping and fracture occurred with sin‐
gle crown (3%) than with their splinted counterparts (0.7%).
To compare the expenses associated with the three groups, one
of the advantages was having a long observational period. To the best
of our knowledge, no study has previously compared the cost of the
three treatment approaches; therefore, a direct comparison of our
findings to the literature was not possible in this regard. As demon‐
strated by the results of the current research, an ISB, due to the
placement of one less implant, presented with a −16% less initial cost
with respect to the other two groups. Additionally, the difference fur‐
ther increased over the study period due to the expenses associated
with treatment of the complications being higher in the NSC groups.
The present study is not exempt from limitations such as dif‐
ferent implant systems, diameters and length; the lack of informa‐
tion about the soft tissue thickness and the bucco‐lingual position
of the examined implants. Therefore, further studies considering
these variables are necessary in the future, to confirm our results.
Additionally, an inherit limitation of the retrospective nature of
this study is the preclusion of identifying the occlusal profile of in‐
dividual patients. In fact, the uncontrollable variability in occlusion
and dynamic of opposing tooth positions poses an inability to gen‐
erate a categorical set of data indicating an exact and persistent
opposing structure, such as tooth only, crown only or both tooth
and crown, across the observational period. For example, a case
may have commenced with an opposing set of teeth at baseline
but along the follow‐up exhibited a loss of one or more of those
teeth. This varying dynamic among cases creates a set of scenarios
that cannot be confined to a fixed set of opposing occlusion cate‐
gories to be tested.
The type of occlusal adjustment implemented is another clini‐
cal parameter essential to ensuring optimal treatment and a more
standardized study population. Within this study's clinical set‐
ting, canine‐guided occlusal correction is the primarily enforced
approach to such cases. However, as with other routine, minute
details of treatment, patient charts did not consistently contain
whether canine‐guided or group function occlusion was em‐
ployed. For that reason, this may be considered a limitation of the
overall study constituents.
In conclusion, an implant‐supported bridge seems to present
the most ideal long‐term therapeutic solution, among the inves‐
tigated approaches in this study, in rehabilitating a 3‐unit eden‐
tulous area by demonstrating: (a) comparable PI rate to NSC and
lower than SC; (b) comparable survival rate to NSC while higher
than SC; (c) similar complication rate to SC while lower than NSC;
(d) higher success rate than both NSC and SC; (e) lower total cost
than NSC and SC.
C O N FL I C T O F I N T E R E S T
The authors do not have any financial interests neither directly nor
indirectly in the companies whose materials were utilized in this
study. This manuscript was partially supported by the University of
Michigan Periodontal Graduate Student Research Fund.
ORCID
Andrea Ravidà  https://orcid.org/0000-0002-3029-8130
Mustafa Tattan  https://orcid.org/0000-0001-7498-8064
Houssam Askar  https://orcid.org/0000-0003-0829-785X
Shayan Barootchi  https://orcid.org/0000-0002-5347-6577
Lorenzo Tavelli  https://orcid.org/0000-0003-4864-3964
Hom‐Lay Wang  https://orcid.org/0000-0003-4238-1799
REFERENCES
Armitage, G. C. (2004). Periodontal diagnoses and classification of
periodontal diseases. Periodontology 2000, 34, 9–21. https://doi.
org/10.1046/j.0906-6713.2002.003421.x
Cannizzaro, G., Felice, P., Soardi, E., Ferri, V., Leone, M., & Esposito, M.
(2011). Immediate loading of 2 (all‐on‐2) versus 4 (all‐on‐4) implants
placed with a flapless technique supporting mandibular cross‐arch
fixed prostheses: Preliminary results from a pilot randomised con‐
trolled trial. European Journal of Oral Implantology, 4, 205–217.
Chrcanovic, B. R., Albrektsson, T., & Wennerberg, A. (2014). Reasons for
failures of oral implants. Journal of Oral Rehabilitation, 41, 443–476.
https://doi.org/10.1111/joor.12157
Clelland, N., Chaudhry, J., Rashid, R. G., & McGlumphy, E. (2016). Split‐
mouth comparison of splinted and nonsplinted prostheses on short
RAVIDÀ et al.
implants: 3‐year results. International Journal of Oral and Maxillofacial
Implants, 31, 1135–1141. https://doi.org/10.11607/jomi.4565
de Souza Batista, V. E., Verri, F. R., Almeida, D. A., Santiago Junior, J. F.,
Lemos, C. A., & Pellizzer, E. P. (2017). Finite element analysis of im‐
plant‐supported prosthesis with pontic and cantilever in the posterior
maxilla. Computer Methods in Biomechanics and Biomedical Engineering,
20, 663–670. https://doi.org/10.1080/10255842.2017.1287905
Eliasson, A., Eriksson, T., Johansson, A., & Wennerberg, A. (2006). Fixed
partial prostheses supported by 2 or 3 implants: A retrospective
study up to 18 years. International Journal of Oral and Maxillofacial
Implants, 21, 567–574.
Faucher, R. R., & Bryant, R. A. (1983). Bilateral fixed splints. International
Journal of Periodontics & Restorative Dentistry, 3, 8–37.
Flemmig, T. F., & Beikler, T. (2009). Decision making in implant dentistry: An
evidence‐based and decision‐analysis approach. Periodontology 2000,
50, 154–172. https://doi.org/10.1111/j.1600-0757.2008.00286.x
Grossmann, Y., Finger, I. M., & Block, M. S. (2005). Indications for splint‐
ing implant restorations. Journal of Oral and Maxillofacial Surgery, 63,
1642–1652. https://doi.org/10.1016/j.joms.2005.05.149
Guichet, D. L., Yoshinobu, D., & Caputo, A. A. (2002). Effect of splint‐
ing and interproximal contact tightness on load transfer by implant
restorations. Journal of Prosthetic Dentistry, 87, 528–535. https://doi.
org/10.1067/mpr.2002.124589
Hasan, I., Bourauel, C., Keilig, L., Stark, H., & Luckerath, W. (2015). The
effect of implant splinting on the load distribution in bone bed around
implant‐supported fixed prosthesis with different framework mate‐
rials: A finite element study. Annals of Anatomy, 199, 43–51. https://
doi.org/10.1016/j.aanat.2014.12.001
Huang, H. L., Huang, J. S., Ko, C. C., Hsu, J. T., Chang, C. H., & Chen,
M. Y. (2005). Effects of splinted prosthesis supported a wide im‐
plant or two implants: A three‐dimensional finite element anal‐
ysis. Clinical Oral Implants Research, 16, 466–472. https://doi.
org/10.1111/j.1600-0501.2005.01124.x
Jung, R. E., Zembic, A., Pjetursson, B. E., Zwahlen, M., & Thoma, D. S.
(2012). Systematic review of the survival rate and the incidence of
biological, technical, and aesthetic complications of single crowns on
implants reported in longitudinal studies with a mean follow‐up of 5
years. Clinical Oral Implants Research, 23(Suppl 6), 2–21.
Malo, P., Nobre, M., & Lopes, A. (2013). Immediate loading of 'all‐on‐4'
maxillary prostheses using trans‐sinus tilted implants without sinus
bone grafting: A retrospective study reporting the 3‐year outcome.
European Journal of Oral Implantology, 6, 273–283.
Malo, P., Rangert, B., & Nobre, M. (2003). "All‐on‐four" immediate‐func‐
tion concept with branemark system implants for completely eden‐
tulous mandibles: A retrospective clinical study. Clinical Implant
Dentistry and Related Research, 5(Suppl 1), 2–9.
Nissan, J., Ghelfan, O., Gross, M., & Chaushu, G. (2010). Analysis of load
transfer and stress distribution by splinted and unsplinted implant‐
supported fixed cemented restorations. Journal of Oral Rehabilitation,
37, 658–662. https://doi.org/10.1111/j.1365-2842.2010.02096.x
Pihlstrom, B. L., Michalowicz, B. S., & Johnson, N. W. (2005). Periodontal
diseases. Lancet, 366, 1809–1820. https://doi.org/10.1016/
S0140-6736(05)67728-8
Pjetursson, B. E., Bragger, U., Lang, N. P., & Zwahlen, M. (2007).
Comparison of survival and complication rates of tooth‐supported
fixed dental prostheses (FDPs) and implant‐supported FDPs and sin‐
gle crowns (SCs). Clinical Oral Implants Research, 18(Suppl 3), 97–113.
https://doi.org/10.1111/j.1600-0501.2007.01439.x
Ravida, A., Barootchi, S., Tattan, M. A., Saleh, M. H. A., Gargallo‐Albiol,
J., & Wang, H. L. (2018). Clinical outcomes and cost effectiveness of
computer‐guided versus conventional implant‐retained hybrid pros‐
theses: A long‐term retrospective analysis of treatment protocols.
Journal of Periodontology, 89, 1015–1024
Ravida, A., Saleh, M. H. A., Muriel, M. C., Maska, B., & Wang, H. L. (2018).
Biological and technical complications of splinted or nonsplinted
|
      305
dental implants: A decision tree for selection. Implant Dentistry, 27,
89–94.
Renvert, S., & Persson, G. R. (2009). Periodontitis as a potential risk fac‐
tor for peri‐implantitis. Journal of Clinical Periodontology, 36(Suppl
10), 9–14. https://doi.org/10.1111/j.1600-051X.2009.01416.x
Schneider, C. A., Rasband, W. S., & Eliceiri, K. W. (2012). NIH Image to
ImageJ: 25 years of image analysis. Nature Methods, 9, 671–675.
https://doi.org/10.1038/nmeth.2089
Serino, G., & Strom, C. (2009). Peri‐implantitis in partially eden‐
tulous patients: Association with inadequate plaque con‐
trol. Clinical Oral Implants Research, 20, 169–174. https://doi.
org/10.1111/j.1600-0501.2008.01627.x
Shigemitsu, R., Ogawa, T., Matsumoto, T., Yoda, N., Gunji, Y., Yamakawa,
Y., … Sasaki, K. (2013). Stress distribution in the peri‐implant bone
with splinted and non‐splinted implants by in vivo loading data‐
based finite element analysis. Odontology, 101, 222–226. https://doi.
org/10.1007/s10266-012-0077-y
Spies, B. C., Witkowski, S., Butz, F., Vach, K., & Kohal, R. J. (2016). Bi‐layered
zirconia/fluor‐apatite bridges supported by ceramic dental implants: A
prospective case series after thirty months of observation. Clinical Oral
Implants Research, 27, 1265–1273. https://doi.org/10.1111/clr.12731
Tonetti, M., Palmer, R. & Working Group 2 of the VIII European
Workshop on Periodontology (2012). Clinical research in implant
dentistry: Study design, reporting and outcome measurements:
Consensus report of working group 2 of the VIII European Workshop
on Periodontology. Journal of Clinical Periodontology, 39(Suppl 12),
73–80. https://doi.org/10.1111/j.1600-051X.2011.01843.x
Vigolo, P., Mutinelli, S., Zaccaria, M., & Stellini, E. (2015). Clinical evalua‐
tion of marginal bone level change around multiple adjacent implants
restored with splinted and nonsplinted restorations: A 10‐year ran‐
domized controlled trial. International Journal of Oral and Maxillofacial
Implants, 30, 411–418. https://doi.org/10.11607/jomi.3837
Wilson, T. G. Jr (2009). The positive relationship between excess cement
and peri‐implant disease: A prospective clinical endoscopic study.
Journal of Periodontology, 80, 1388–1392. https://doi.org/10.1902/
jop.2009.090115
Yi, Y. J., Lee, J. Y., & Kim, Y. K. (2013). Comparative clinical study of three‐
unit fixed partial prostheses supported by two or three implants.
International Journal of Oral and Maxillofacial Implants, 28, 1110–1115.
https://doi.org/10.11607/jomi.2940
Yilmaz, B., Seidt, J. D., McGlumphy, E. A., & Clelland, N. L. (2011).
Comparison of strains for splinted and nonsplinted screw‐retained
prostheses on short implants. International Journal of Oral and
Maxillofacial Implants, 26, 1176–1182.
Yoda, N., Liao, Z., Chen, J., Sasaki, K., Swain, M., & Li, Q. (2016). Role
of implant configurations supporting three‐unit fixed partial den‐
ture on mandibular bone response: Biological‐data‐based finite el‐
ement study. Journal of Oral Rehabilitation, 43, 692–701. https://doi.
org/10.1111/joor.12411
S U P P O R T I N G I N FO R M AT I O N
Additional supporting information may be found online in the
Supporting Information section at the end of the article. 
How to cite this article: Ravidà A, Tattan M, Askar H, Barootchi
S, Tavelli L, Wang H‐L. Comparison of three different types of
implant-supported fixed dental prostheses: A long‐term
retrospective study of clinical outcomes and cost‐
effectiveness. Clin Oral Impl Res. 2019;30:295–305. https://doi.
org/10.1111/clr.13415